Your proposal is intriguing, and I'm always in favor of seeking win-win situations.
But ... I suspect that legislators who currently have the power to gerrymander outrageously won't be willing to settle for the power to gerrymander just a little bit.
I suspect that they will gladly fight in court to preserve their power. And if somehow they're forced to use independent commissions, they'll do their best to stack those. And they won't worry about the possibility that sometime in the future the other side will prevail and gerrymander against them.
(Did Mitch McConnell hold back on stacking the Supreme Court because he worried that the Democrats would someday be in power and do the same? No---he used his power when he had it.)
In other words, won't the legislators who are in the majority see this as a lose-win situation for them?
I’m going to give you two answers for two different scenarios:
#1: Congress passes a federal law that governs redistricting using this approach.
#2: State legislatures pass state laws governing redistricting using this approach.
Which is relevant would depend on how this plays out.
Consider #1. Congress is almost evenly divided now, although the Senate doesn’t behave that way because of the filibuster. Representatives and Senators don’t hold the redistricting power now. But, of course, they’re all interested in advantaging their respective parties. Until recently, gerrymandering clearly favored the Republicans. But two things have changed recently: (1) Democrats have won battles in state courts to overturn some of the worst (i.e., most effective) Republican-leaning gerrymanders; and (2) Democrats have become more aggressive and effective in creating Democratic-leaning gerrymanders. So, gerrymandering no longer strongly benefits one party over the other at the national level.
Interestingly, in today’s NY Times (https://www.nytimes.com/2022/03/10/us/politics/redistricting-congressional-maps-elections.html), Nate Cohn analyzes congressional maps and concludes “In a departure from a decades-long pattern in American politics, this year’s national congressional map is poised to be balanced between the two parties, with a nearly equal number of districts that are expected to lean Democratic and Republican for the first time in more than 50 years.”
Recognizing this new reality, I think that there’s some hope that Congress might be willing to take a step like I’m suggesting even though Republicans have opposed the *much* broader The Freedom to Vote Act of 2021.
Certainly not a slam dunk, but plausible. And, SCOTUS is unlikely to block such a law: In the Rucho v. Common Cause decision, Justice Roberts’ majority decision said “Aware of electoral districting problems, the Framers chose a characteristic approach, assigning the issue to the state legislatures, expressly checked and balanced by the Federal Congress, with no suggestion that the federal courts had a role to play.”
Consider #2. Legislators of both parties are being taught several lessons. First, extreme gerrymandering is causing their power to be taken away in states where ballot initiatives are allowed, showing that voters dislike gerrymandering. Second, although SCOTUS has refused to take action, state supreme courts have been taking action based on state constitutional issues and, so far, SCOTUS has declined to overrule the state supreme court actions. Third, Democrats are doing more aggressive gerrymanders and they’re getting away with some of it.
So, at the state level, it is possible that legislators might see the writing on the wall that gerrymandering is less one-sided than it used to be and that just because they’re in power now doesn’t mean that they’ll always be in power.
Summary: Approach #1 is preferred. We already know that 48 Democratic Senators support the more overarching Freedom to Vote Act and that all Republican Senators oppose it. But I wonder/hope that 12 Republicans could be found that would support this much narrower proposal, especially if they could be convinced that this would protect Republicans as well as Democrats and that voters want it.
The biggest problem I see is how to build any interest/momentum for such an approach. I don’t have a good answer to that.
Yes. In fact, point #2 in the section titled "How Could This Work?" proposes a criterion: "The [proposed] map would be evaluated against the ensemble. The map would be acceptable if the map’s “dot” on the box plot for each district falls within the box, meaning that the map defines districts that are most typical — they fall within the middle 50% of each district’s ensemble histogram. This requirement would allow some bias in the map, but would prohibit egregious gerrymandering like we’ve seen in the maps in North Carolina and other states."
There could be some debate about whether this criterion is strict enough or loose enough. As an example, the judges' map would not quite meet this criterion. To achieve a win-win, it is important to choose a criterion that would allow some limited ability for the legislature to shape the map to meet their goal, while limiting the amount of partisan bias that is acceptable. I believe that this 50% criterion would be a good balance, but it would be good to get input from the experts.
If a map falls “in the middle 50% of each district’s ensemble histogram” it will be quite unbiased. In fact, that might even be too tight a constraint. For example, the hypothetical map produced by the non-partisan group of judges would not pass muster by that criterion. The judges’ map was much better than the official map, but still leaned Republican. If a proposal like this were to get traction, this is an area where some more input by the experts would be very helpful.
The "win-win" idea sounds resonable, but as always "the devil is in the details" and I am skeptical that legislatures (in either NY or NC) could agree on how to construct reasonable "ensembles" by which to evaluate newly drawn maps. So, really nice idea, but I doubt it will work. Now a couple of specific issues. You say "With the Court’s recent conservative shift, it is unclear how the Court would rule on a similar suit today." about
with Kavanaugh replacing Kennedy and Barrett replacing Ginsburg, I don't think the current court would be other than 6-3 the other way ... do you?
I read the first part of the Bangia et al article from which you included the right half of their figure 5. While the figure supports your (and their) argument, it (and the left half of that figure) have an "artifact" which, IMHO, deserves explanation but has none. Bangia et al note that "the distribution of the 6th most Republican distict is quite peaked" (in both halves of the figure) but do not comment on the fact that the same peak, at the same democratic vote fraction is there in the 5th and 7th (and to a lesser degree in the 4th) most Republican districts (MRDs). I find this quite surprising and to me it deserves an explanation. The histograms for the 2-4 and 8-10 MRDs have a "sort of" unimodal distribution resembling your "schematic view". To me it is not that surprising the extreme districts have weird histograms, but that the MRDs 5-7 have this weird peak, while MRDs 2-4 and 8-10 do not makes me wonder what is going on.
This is to reply to your comment about the peaked distributions. First of all, good question!
I'm not certain, but here's my theory: In the 2016 election, parts of the state voted in the 60-69% range for Republicans, parts voted in the 50-59% range for Republican, and parts in the 60-69% range for Democrats. (See the summary map in the article at https://en.wikipedia.org/wiki/2016_United_States_House_of_Representatives_elections_in_North_Carolina.) Since the districts you're discussing were neither MRDs nor MDDs (most Democratic districts), they're likely to be in the 50-59% range for Republicans, which (ignoring votes for third party candidates) is the 41-50% range for Democrats. So you would expect that for those "middle" districts the distributions would have peaks in the low-to-mid 40s, which is what you see.
First let me comment that my cavil about the distributions may not bother other people ... I am sorta into statistics. Even so, I think the "box plots" are just as pursasive and do not have this distration. I agree that the "middle" districts (neither MRD nor MDD) are likely to have "in the middle" outcomes. It is the sharpness of the peaks and the fact that they are located at (apparently) exactly the same Democratic vote fraction for 3 or 4 (contiguous) "middlish" districts. That specific Democratic vote fraction is lower in 2016 than 2012, but it still seems strange. Bangia et al mention one district (the 6th) where the sharp peak (approximately) corresponds with the median ... but the same sharp peak is there at the same spot in the next more and less Republican districts where it is above or below the median. And in some way, to me at least, it detracts from the point. If districts 5-7 looked more like 3 and 9: sort of smooth and gaussian-like, the contrast with the 2 or 3 MDDs would be much greater. As I said in my original comment I would be less surprised with "weird looking" distribution for the extreme (MDD, MRD) districts than those in the middle.
I agree that the box plots are just as pervasive and don't have this distraction. In fact, in the Bangia et al. article, they say something along those lines. I included them in the my description because I felt that even people not into statistics have seen distributions before, so might be able to develop some intuition about what is going on than if they just saw the box plots.
I think that the MDD and MRD districts don't look "weird" (to use your term) because there are a wide range of maps in the ensemble for which those districts would remain MDD or MRD. I wish could be more precise about this, but I can't.
As I said I'm interested in statistics, and this sort of thing in particular: constructing a "large" number of synthetic data sets and using them to evaluate results from analyses of real data. I put large in quotes because I published a paper in a respectable professional journal more than 45 years ago where the number of synthetic data sets we generated was ... ta da ... 100. Seeing those spikes in the results from thousands of computer generated districts just really caught my eye! https://ui.adsabs.harvard.edu/abs/1976ApJ...203..286K/abstract
This is to reply to your comment about my statement that "With the Court’s recent conservative shift, it is unclear how the Court would rule on a similar suit today." I agree with you that if the Court accepted the case, the ruling would be 6-3 the other way. What I'm not sure of is whether the Court would accept a similar challenge.
I'd guess that they would but I don't understand enough about how those decisions are made to feel comfortable making a prediction.
If I was a legislative leader trying to maintain the power to gerrymander, I'd certainly be trying to shut down independent commissions and bringing such a suit seems like a good option given the Court's current composition.
That's one of the reasons (but not the only one) that I'm skeptical that independent commissions will solve partisan gerrymandering.
I'm glad I got your attention Josh. Since you raise a number of points and issues, I'm going to address them one at a time in separate replies because I think that will make it easier for others to follow.
I'll start with your skepticism that legislatures could agree on how to construct reasonable ensembles by which to evaluate newly drawn maps. I share your skepticism that legislatures could agree on how to construct reasonable ensembles. Anything that requires legislators to agree on what is reasonable is doomed to fail.
Fortunately, that's not how the ensemble method works. The ensemble is a sampling of the distribution of all maps that meet the state's *legal* requirements. (Examples of legal requirements include that each district is contiguous, that all districts contain the same population within some tolerance, that counties shouldn't be split across districts if possible, and that the map meets the requirements of the Voting Rights Act.) No judgement of reasonableness goes into creating the ensemble. That's crucial because, as you point out, you'd never get legislators to agree on what is reasonable or not.
Moreover, there's never a need for anyone to decide whether or not a map is reasonable. The method just decides whether or not a particular map is a statistical rarity in the ensemble, that is, that the particular map is extremely unlikely to occur.
The reason that this is important is that in court cases the gerrymanderers argue "that's just the way the votes worked out" but this method shows that the gerrymandered maps are exceedingly unlikely to occur by chance (just like an event from a normal distribution is exceedingly unlikely to occur 3 standard deviations from the mean).
Does this make you feel more optimistic about whether my proposed win-win could work? If not, why not? If so, any suggestions on how I could make this more clear?
Thanks for the clarification. It is encouraging that the NC courts have accepted the ensemble analysis as credible expert evidence in their "fact finding", but I note that the NC supreme court decision was split along party lines. I am still skeptical that a (Republican dominated for the moment) legislature would "trust" the technology to be neutral/non-partisan in the construction of the ensembles ... Republicans of late have tended to opt for their own "alternative facts". I think I do understand how the ensemble method for estimating the likelihood that (say) a particular map gave one party a big advantage "by chance", and you explained it well.
Lee,
Your proposal is intriguing, and I'm always in favor of seeking win-win situations.
But ... I suspect that legislators who currently have the power to gerrymander outrageously won't be willing to settle for the power to gerrymander just a little bit.
I suspect that they will gladly fight in court to preserve their power. And if somehow they're forced to use independent commissions, they'll do their best to stack those. And they won't worry about the possibility that sometime in the future the other side will prevail and gerrymander against them.
(Did Mitch McConnell hold back on stacking the Supreme Court because he worried that the Democrats would someday be in power and do the same? No---he used his power when he had it.)
In other words, won't the legislators who are in the majority see this as a lose-win situation for them?
I’m going to give you two answers for two different scenarios:
#1: Congress passes a federal law that governs redistricting using this approach.
#2: State legislatures pass state laws governing redistricting using this approach.
Which is relevant would depend on how this plays out.
Consider #1. Congress is almost evenly divided now, although the Senate doesn’t behave that way because of the filibuster. Representatives and Senators don’t hold the redistricting power now. But, of course, they’re all interested in advantaging their respective parties. Until recently, gerrymandering clearly favored the Republicans. But two things have changed recently: (1) Democrats have won battles in state courts to overturn some of the worst (i.e., most effective) Republican-leaning gerrymanders; and (2) Democrats have become more aggressive and effective in creating Democratic-leaning gerrymanders. So, gerrymandering no longer strongly benefits one party over the other at the national level.
Interestingly, in today’s NY Times (https://www.nytimes.com/2022/03/10/us/politics/redistricting-congressional-maps-elections.html), Nate Cohn analyzes congressional maps and concludes “In a departure from a decades-long pattern in American politics, this year’s national congressional map is poised to be balanced between the two parties, with a nearly equal number of districts that are expected to lean Democratic and Republican for the first time in more than 50 years.”
Recognizing this new reality, I think that there’s some hope that Congress might be willing to take a step like I’m suggesting even though Republicans have opposed the *much* broader The Freedom to Vote Act of 2021.
Certainly not a slam dunk, but plausible. And, SCOTUS is unlikely to block such a law: In the Rucho v. Common Cause decision, Justice Roberts’ majority decision said “Aware of electoral districting problems, the Framers chose a characteristic approach, assigning the issue to the state legislatures, expressly checked and balanced by the Federal Congress, with no suggestion that the federal courts had a role to play.”
Consider #2. Legislators of both parties are being taught several lessons. First, extreme gerrymandering is causing their power to be taken away in states where ballot initiatives are allowed, showing that voters dislike gerrymandering. Second, although SCOTUS has refused to take action, state supreme courts have been taking action based on state constitutional issues and, so far, SCOTUS has declined to overrule the state supreme court actions. Third, Democrats are doing more aggressive gerrymanders and they’re getting away with some of it.
So, at the state level, it is possible that legislators might see the writing on the wall that gerrymandering is less one-sided than it used to be and that just because they’re in power now doesn’t mean that they’ll always be in power.
Summary: Approach #1 is preferred. We already know that 48 Democratic Senators support the more overarching Freedom to Vote Act and that all Republican Senators oppose it. But I wonder/hope that 12 Republicans could be found that would support this much narrower proposal, especially if they could be convinced that this would protect Republicans as well as Democrats and that voters want it.
The biggest problem I see is how to build any interest/momentum for such an approach. I don’t have a good answer to that.
I would agree with this if the % to which the maps are gerrymandered would be limited to a small %. What would you recommend?
Yes. In fact, point #2 in the section titled "How Could This Work?" proposes a criterion: "The [proposed] map would be evaluated against the ensemble. The map would be acceptable if the map’s “dot” on the box plot for each district falls within the box, meaning that the map defines districts that are most typical — they fall within the middle 50% of each district’s ensemble histogram. This requirement would allow some bias in the map, but would prohibit egregious gerrymandering like we’ve seen in the maps in North Carolina and other states."
There could be some debate about whether this criterion is strict enough or loose enough. As an example, the judges' map would not quite meet this criterion. To achieve a win-win, it is important to choose a criterion that would allow some limited ability for the legislature to shape the map to meet their goal, while limiting the amount of partisan bias that is acceptable. I believe that this 50% criterion would be a good balance, but it would be good to get input from the experts.
I also looked at the analysis (https://sites.duke.edu/quantifyinggerrymandering/files/2021/11/congressionalReport_Comparison.pdf) that Prof. Mattingly and his team did on the original 2021 maps for the NC US Congressional districts: it is very far away from meeting this criterion; likewise for the 2021 maps for NCGA Senate (https://sites.duke.edu/quantifyinggerrymandering/files/2021/11/senateNCGAReport_comparison.pdf) and House (https://sites.duke.edu/quantifyinggerrymandering/files/2021/11/houseNCGAReport_Comparison.pdf).
For us not as mathematical, it is unclear how biased it could be if it falls "in the middle 50% of each district's ensemble histogram".
If a map falls “in the middle 50% of each district’s ensemble histogram” it will be quite unbiased. In fact, that might even be too tight a constraint. For example, the hypothetical map produced by the non-partisan group of judges would not pass muster by that criterion. The judges’ map was much better than the official map, but still leaned Republican. If a proposal like this were to get traction, this is an area where some more input by the experts would be very helpful.
Thanks Lee
The "win-win" idea sounds resonable, but as always "the devil is in the details" and I am skeptical that legislatures (in either NY or NC) could agree on how to construct reasonable "ensembles" by which to evaluate newly drawn maps. So, really nice idea, but I doubt it will work. Now a couple of specific issues. You say "With the Court’s recent conservative shift, it is unclear how the Court would rule on a similar suit today." about
https://www.scotusblog.com/case-files/cases/arizona-state-legislature-v-arizona-independent-redistricting-commission/
with Kavanaugh replacing Kennedy and Barrett replacing Ginsburg, I don't think the current court would be other than 6-3 the other way ... do you?
I read the first part of the Bangia et al article from which you included the right half of their figure 5. While the figure supports your (and their) argument, it (and the left half of that figure) have an "artifact" which, IMHO, deserves explanation but has none. Bangia et al note that "the distribution of the 6th most Republican distict is quite peaked" (in both halves of the figure) but do not comment on the fact that the same peak, at the same democratic vote fraction is there in the 5th and 7th (and to a lesser degree in the 4th) most Republican districts (MRDs). I find this quite surprising and to me it deserves an explanation. The histograms for the 2-4 and 8-10 MRDs have a "sort of" unimodal distribution resembling your "schematic view". To me it is not that surprising the extreme districts have weird histograms, but that the MRDs 5-7 have this weird peak, while MRDs 2-4 and 8-10 do not makes me wonder what is going on.
As you can tell you got my attention.
This is to reply to your comment about the peaked distributions. First of all, good question!
I'm not certain, but here's my theory: In the 2016 election, parts of the state voted in the 60-69% range for Republicans, parts voted in the 50-59% range for Republican, and parts in the 60-69% range for Democrats. (See the summary map in the article at https://en.wikipedia.org/wiki/2016_United_States_House_of_Representatives_elections_in_North_Carolina.) Since the districts you're discussing were neither MRDs nor MDDs (most Democratic districts), they're likely to be in the 50-59% range for Republicans, which (ignoring votes for third party candidates) is the 41-50% range for Democrats. So you would expect that for those "middle" districts the distributions would have peaks in the low-to-mid 40s, which is what you see.
First let me comment that my cavil about the distributions may not bother other people ... I am sorta into statistics. Even so, I think the "box plots" are just as pursasive and do not have this distration. I agree that the "middle" districts (neither MRD nor MDD) are likely to have "in the middle" outcomes. It is the sharpness of the peaks and the fact that they are located at (apparently) exactly the same Democratic vote fraction for 3 or 4 (contiguous) "middlish" districts. That specific Democratic vote fraction is lower in 2016 than 2012, but it still seems strange. Bangia et al mention one district (the 6th) where the sharp peak (approximately) corresponds with the median ... but the same sharp peak is there at the same spot in the next more and less Republican districts where it is above or below the median. And in some way, to me at least, it detracts from the point. If districts 5-7 looked more like 3 and 9: sort of smooth and gaussian-like, the contrast with the 2 or 3 MDDs would be much greater. As I said in my original comment I would be less surprised with "weird looking" distribution for the extreme (MDD, MRD) districts than those in the middle.
I agree that the box plots are just as pervasive and don't have this distraction. In fact, in the Bangia et al. article, they say something along those lines. I included them in the my description because I felt that even people not into statistics have seen distributions before, so might be able to develop some intuition about what is going on than if they just saw the box plots.
I think that the MDD and MRD districts don't look "weird" (to use your term) because there are a wide range of maps in the ensemble for which those districts would remain MDD or MRD. I wish could be more precise about this, but I can't.
As I said I'm interested in statistics, and this sort of thing in particular: constructing a "large" number of synthetic data sets and using them to evaluate results from analyses of real data. I put large in quotes because I published a paper in a respectable professional journal more than 45 years ago where the number of synthetic data sets we generated was ... ta da ... 100. Seeing those spikes in the results from thousands of computer generated districts just really caught my eye! https://ui.adsabs.harvard.edu/abs/1976ApJ...203..286K/abstract
This is to reply to your comment about my statement that "With the Court’s recent conservative shift, it is unclear how the Court would rule on a similar suit today." I agree with you that if the Court accepted the case, the ruling would be 6-3 the other way. What I'm not sure of is whether the Court would accept a similar challenge.
I'd guess that they would but I don't understand enough about how those decisions are made to feel comfortable making a prediction.
If I was a legislative leader trying to maintain the power to gerrymander, I'd certainly be trying to shut down independent commissions and bringing such a suit seems like a good option given the Court's current composition.
That's one of the reasons (but not the only one) that I'm skeptical that independent commissions will solve partisan gerrymandering.
I agree that independent commissions are not likely to help, particularly with the current make up of the SCOTUS.
I'm glad I got your attention Josh. Since you raise a number of points and issues, I'm going to address them one at a time in separate replies because I think that will make it easier for others to follow.
I'll start with your skepticism that legislatures could agree on how to construct reasonable ensembles by which to evaluate newly drawn maps. I share your skepticism that legislatures could agree on how to construct reasonable ensembles. Anything that requires legislators to agree on what is reasonable is doomed to fail.
Fortunately, that's not how the ensemble method works. The ensemble is a sampling of the distribution of all maps that meet the state's *legal* requirements. (Examples of legal requirements include that each district is contiguous, that all districts contain the same population within some tolerance, that counties shouldn't be split across districts if possible, and that the map meets the requirements of the Voting Rights Act.) No judgement of reasonableness goes into creating the ensemble. That's crucial because, as you point out, you'd never get legislators to agree on what is reasonable or not.
Moreover, there's never a need for anyone to decide whether or not a map is reasonable. The method just decides whether or not a particular map is a statistical rarity in the ensemble, that is, that the particular map is extremely unlikely to occur.
The reason that this is important is that in court cases the gerrymanderers argue "that's just the way the votes worked out" but this method shows that the gerrymandered maps are exceedingly unlikely to occur by chance (just like an event from a normal distribution is exceedingly unlikely to occur 3 standard deviations from the mean).
Does this make you feel more optimistic about whether my proposed win-win could work? If not, why not? If so, any suggestions on how I could make this more clear?
Thanks for the clarification. It is encouraging that the NC courts have accepted the ensemble analysis as credible expert evidence in their "fact finding", but I note that the NC supreme court decision was split along party lines. I am still skeptical that a (Republican dominated for the moment) legislature would "trust" the technology to be neutral/non-partisan in the construction of the ensembles ... Republicans of late have tended to opt for their own "alternative facts". I think I do understand how the ensemble method for estimating the likelihood that (say) a particular map gave one party a big advantage "by chance", and you explained it well.